基于近红外光谱的面粉营养成分分析实验数据

小麦面粉是烹饪和烘焙中一种非常重要的食材，具有许多重要的特性和功能。面粉中含有丰富的碳水化合物、蛋白质、维生素和矿物质等，是人类日常饮食中重要的能量来源。常见的面粉种类包括普通面粉、全麦面粉、低筋面粉、中筋面粉、高筋面粉、自发面粉等，由于不同面粉的种类也与其适用的用途和配方有关，所以需要有一种快速高效的鉴别方法。近红外光谱法具有快速、方便、无损、算法规则简要说明高效等优点。将化学计量法结合近红外光谱对面粉进行识别是一种比较常见的方法。通过对实验数据的分析总结，可对市场主流面粉进行快速分析，获得面粉的脂肪、蛋白质、灰质、水分等营养成分数值。采用适当的模型，采用端到端的训练方式，通过原始近红外光谱数据即可得到分类结果，简化了近红外光谱分析的流程，同时将经过训练的模型部署在嵌入式设备中，有助于实现近红外光谱仪的智能化。

Wheat flour is a critically important ingredient in cooking and baking, with numerous significant properties and functions. It is rich in carbohydrates, proteins, vitamins, minerals and other nutrients, making it a vital energy source in human daily diets. Common types of flour include all-purpose flour, whole wheat flour, low-gluten flour, medium-gluten flour, high-gluten flour, self-raising flour, and so on. Since different flour types are associated with specific applicable scenarios and recipes, there is an urgent need for a rapid and efficient identification method. Near-infrared spectroscopy (NIRS) has advantages such as rapidity, convenience, non-destructiveness, simple algorithmic requirements and high efficiency. Combining chemometrics with NIRS for flour identification is a relatively common approach. Through analysis and summarization of experimental data, rapid analysis can be performed on mainstream commercial flours to obtain the contents of nutrient components including fat, protein, ash and moisture. By adopting appropriate models and end-to-end training methods, classification results can be directly derived from raw near-infrared spectral data, which simplifies the workflow of near-infrared spectroscopic analysis. Additionally, deploying the trained models on embedded devices can help realize the intellectualization of near-infrared spectrometers.